CN105809641B - The exposure compensating and edge enhancing method of a kind of mist elimination image - Google Patents

Abstract

The invention discloses a kind of exposure compensating of mist elimination image and edge enhancing method, a kind of improved colour killing algorithm based on significance guiding is combined with gamma transformation, brightness layer to mist elimination image is modified compensation, the problem of so as to effectively improve mist elimination image under-exposure；And the method by carrying out marginal information compensation to the input item of Steerable filter device, the problem of effectively improving mist elimination image edge blurry.

Description

Exposure compensation and edge enhancement method for defogged image

Technical Field

The invention belongs to the field of digital image processing, and particularly relates to an exposure compensation and edge enhancement method for a defogged image.

Background

Images and videos shot in foggy weather or dust-haze weather often have the phenomenon of poor image quality and low definition. In order to improve the quality and the definition of the image, the image needs to be defogged.

The image defogging method based on dark channel priority (dark channel prior) is one of the defogging methods based on the fog model, and achieves better effect in practice. Since the transmittance estimation is low in this method and needs to be optimized by a soft-matting method, the main disadvantages are high computational complexity and dark defogged images, i.e., underexposure. In order to increase the calculation speed, the coarse transmittance estimation can be optimized by using a guiding filter, but in the case that the guiding filter is a fog image with a blurred guiding image, the obtained defogged image generates an edge blurring phenomenon.

In view of the above analysis, the major drawbacks of the method using a combination of dark channel precedence and guided filters (i.e., GDCP defogging algorithm) are manifested as underexposure and edge blurring.

Disclosure of Invention

In view of the above, the present invention provides an exposure compensation and edge enhancement method for a defogged image, which can effectively improve the problems of underexposure and edge blurring of the defogged image.

The technical scheme for realizing the invention is as follows:

an exposure compensation and edge enhancement method of a defogged image comprises the following steps:

step one, calculating a dark channel I of a fog image I to be processed according to the definition of the dark channel priority^dark(ii) a And according to the dark channel I^darkEstimating atmospheric light A;

step two, calculating a dark channel of the I/A according to the atmospheric light A obtained by estimation; estimation of initial transmission from dark channels of I/AFor initial transmittancePerforming Gaussian smoothing to obtain smooth transmittance t^s；

Step three, solving the Laplacian pyramid L of the 0 th layer of the fog image I by adopting an image pyramid technology₀And using the formula (3) for the smooth transmittance t^sOptimizing to obtain t^e：

Wherein,indicating guided filtering, ξ is a preset adjustmentThe ratio parameter of (a) to (b),is the Laplacian pyramid L of the 0 th layer of the fog image I₀The luminance image of (a);

step four, obtaining the atmospheric light A and t according to estimation^eCalculate the initial defogged image J^L：

Step five, decomposing an initial defogged image J^LAnd a luminance layer L, calculating to obtain a luminance image according to the luminance layer LTo pairStandardizing and fusing to obtain achromatic image I_dec：

Step six, canceling image I_decPerforming gamma conversion to obtain L';

step seven, maintaining the initial defogged image J^LAnd (3) replacing the brightness layer L of the image with the L' to reconstruct the final defogged image.

Further, the method for estimating the atmospheric light a comprises: taking a dark channel I of a fog image I to be processed^darkTaking the pixel set B with the maximum brightness of 0.1%, and then taking the pixel set B corresponding to the pixel set B in the original image I_IFrom B_IAnd selecting three channels of RGB and the largest pixel point as the estimation of A.

Further, the initial transmittanceThe estimation method comprises the following steps:

where ω is a constant close to 1 but less than 1, c is one of the color channels { r, g, b } of the fog image I; Ω is a local window area centered on x, y denotes the pixels in the Ω window, I^c(y) represents the value of the c-th color layer of the fog image I at y, A^cAnd the value of the c color layer of A is shown.

Further, the luminance image is obtained through calculation according to the luminance layer LThe method specifically comprises the following steps:

wherein G [. C]Is a gaussian feathering operator; m_mask(x) Represents M_maskTaking a value at the point x; wherein S (x) is not less than mu and L (x) is not less than v; h is a chrominance layer, S is a saturation layer; the parameter k is a preset period adjustment parameter,is a compensation angle, τ is a preset parameter for reducing the effect of saturation; μ and v are preset thresholds; n is the number of pixels in the region Ω, s (x) represents the saturation at point x, and l (x) represents the luminance at point x.

Further, theThe standardization and fusion processing methods are respectively as follows:

wherein,for normalizing luminance image, χ is a predetermined parameter for adjusting global intensity to avoid overexposure or underexposure of the image, η is a predetermined control L andthe mixing ratio parameter of (1).

Has the advantages that:

the invention provides an improved algorithm aiming at exposure compensation and edge enhancement of a GDCP defogging algorithm. In the invention, an improved achromatism algorithm based on saliency guidance is combined with gamma conversion to correct and compensate the brightness layer of the defogged image, thereby effectively improving the problem of insufficient exposure of the defogged image; and the problem of edge blurring of the defogged image is effectively improved by a method of performing edge information compensation on the input item of the guide filter.

Drawings

FIG. 1 is a schematic flow chart of an image defogging method according to the present invention.

Detailed Description

The invention is described in detail below by way of example with reference to the accompanying drawings.

As shown in fig. 1, the present invention provides an exposure compensation and edge enhancement method for a defogged image, comprising the steps of:

step one, calculating a dark channel I of a fog image I to be processed according to the definition of the dark channel priority^dark(ii) a And according to the dark channel I of the fog image I^darkEstimating atmospheric light A;

the calculation formula for the dark channel can be expressed as:

wherein c is one of the color channels { r, g, b } of the fog image I; Ω is a local window area centered on x, and y represents the pixels in the Ω window; i is^dark(x) I.e. the dark channel map of i (x). I is^c(y) represents the value of the c-th color layer of the fog image I at y.

The method for estimating the atmospheric light A comprises the following steps: taking a dark channel I of a fog image I to be processed^darkTaking the pixel set B with the maximum brightness of 0.1%, and then taking the pixel set B corresponding to the pixel set B in the original image I_IFrom B_IAnd selecting three channels of RGB and the largest pixel point as the estimation of A.

Step two, calculating a dark channel of the I/A according to the atmospheric light A obtained by estimation; estimation of initial transmission from dark channels of I/AFor initial transmittancePerforming Gaussian smoothing to obtain smooth transmittance t^s；

Initial transmittanceThe calculation formula of (2) is as follows:

where ω is a constant near 1 but less than 1. At this time, ω is equivalent to moderately preserving a certain depth of field for distant scenes. In general, ω is preferably set to 0.95. A. the^cAnd the value of the c color layer of A is shown.

Due to initial transmittance of the imageThere is a block effect, here on the initial transmittance, in order to reduce the transfer of information at the edges of these blocksAnd performing Gaussian smoothing processing.

Step three, solving the Laplacian pyramid L of the 0 th layer of the fog image I by adopting an image pyramid technology₀And using the formula (3) for the smooth transmittance t^sOptimizing to obtain t^e：

Wherein,indicating guided filtering, ξ is a preset adjustmentξ in the present invention takes the value of 1.Is the Laplacian pyramid L of the 0 th layer of the fog image I₀Luminance image of (1), i.e. L₀Average value of each color channel.

In fact, the root cause of the haze image blur caused by the GDCP defogging algorithm is that in the optimization process of transmittance by using the guide filter, the guide image is the fog image i (x), and i (x) is blurred and the edge information is not obvious enough. Therefore, the Laplacian pyramid capable of reflecting the edge information is added to the input item of the guide filter to compensate the edge information. Here, the luminance image L of the 0 th layer laplacian pyramid is selected₀Because of the need for and smooth transmittance t^sThe resolution of (c) is the same.

Since the guiding filter is a local linear filter, the transmittance image after enhancing details can be expressed as:

wherein,

this means that the steering filter needs to be computed only once, and thus does not add computational complexity while edge enhancement.

Step four, obtaining the atmospheric light A and t according to estimation^eCalculate the initial defogged image J^L：

The calculation formula is as follows:

step five, decomposing an initial defogged image J^LThe luminance layer L of (1), the luminance image is obtained according to the formulas (6) and (7)To pairStandardizing and fusing to obtain achromatic image I_dec：

Wherein G [. C]Is a gaussian feathering operator; m_mask(x) Represents M_maskTaking a value at the point x; wherein S (x) is not less than mu and L (x) is not less than v; h is a chrominance layer, S is a saturation layer; the parameter k is a preset periodic adjustment parameter, which can maintain the opposite color,is a compensation angle, tau is a preset parameter for reducing the influence of saturation, and the function is equivalent to controlling a chroma contrast modulator; μ and v are preset thresholds used to filter discontinuities in the salient region; in the present invention, the default values of these two parameters are: mu is 0.1, v is 0.6

N is the number of pixels in the region Ω, s (x) represents the saturation at point x, and l (x) represents the luminance at point x;

wherein,for normalizing luminance images, the parameter χ is used to adjust global intensity to avoid overexposure or underexposure of the image, η is the sum of controls L andthe mixing ratio parameter of (1).

One drawback of the saliency-based guided achromatic algorithm is that when smooth bright areas exist in the image, the computed achromatic image has discontinuities. Here, fusing these two discontinuous regions can effectively eliminate the discontinuous edges.

Step six, canceling image I_decPerforming gamma conversion to obtain L';

step seven, maintaining the initial defogged image J^LAnd (3) replacing the brightness layer L of the image with the L' to reconstruct the final defogged image.

And from the fifth step to the seventh step, exposure compensation is carried out on the initial defogged image. The conventional exposure compensation method is to apply a simple gamma transform to the defogged image. As a result of this process, although the overall brightness of the image is improved, the chrominance contrast of the image is affected. According to the exposure compensation method, before gamma conversion is adopted, the chroma of an original defogged image is enhanced by using a decoloring algorithm based on saliency guidance. Thus, the image after exposure compensation can better maintain the chroma contrast of the original image.

In summary, the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (3)

1. An exposure compensation and edge enhancement method for a defogged image, comprising the steps of:

step one, calculating a dark channel I of a fog image I to be processed^dark(ii) a And according to the dark channel I^darkEstimating atmospheric light A;

step two, calculating a dark channel of the I/A according to the atmospheric light A obtained by estimation; estimation of initial transmission from dark channels of I/AFor initial transmittancePerforming Gaussian smoothing to obtain smooth transmittance t^s；

Step three, solving the Laplacian pyramid L of the 0 th layer of the fog image I by adopting an image pyramid technology₀And using the formula (3) for the smooth transmittance t^sOptimizing to obtain t^e：

Wherein,indicating guided filtering, ξ is a preset adjustmentThe ratio parameter of (a) to (b),is the Laplacian pyramid L of the 0 th layer of the fog image I₀The luminance image of (a);

step four, obtaining the atmospheric light A and t according to estimation^eCalculate the initial defogged image J^L：

Step five, decomposing an initial defogged image J^LAnd a luminance layer L, calculating to obtain a luminance image according to the luminance layer LTo pairStandardizing and fusing to obtain achromatic image I_dec：

The brightness image is obtained by calculation according to the brightness layer LThe method specifically comprises the following steps:

<mrow> <msub> <mi>M</mi> <mrow> <mi>m</mi> <mi>a</mi> <mi>s</mi> <mi>k</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>G</mi> <mo>&amp;lsqb;</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mn>1</mn> </mtd> <mtd> <mrow> <mi>S</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>&amp;le;</mo> <mi>&amp;mu;</mi> <mi>a</mi> <mi>n</mi> <mi>d</mi> <mi> </mi> <mi>L</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>&amp;GreaterEqual;</mo> <mi>v</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <mi>o</mi> <mi>t</mi> <mi>h</mi> <mi>e</mi> <mi>r</mi> <mi>w</mi> <mi>i</mi> <mi>s</mi> <mi>e</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>&amp;rsqb;</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>6</mn> <mo>)</mo> </mrow> </mrow>

wherein G [. C]Is a gaussian feathering operator; m_mask(x) Represents M_maskTaking a value at the point x; wherein S (x) is not less than mu and L (x) is not less than v; h is a chrominance layer, S is a saturation layer; the parameter k is a preset period adjustment parameter,is a compensation angle, τ is a preset parameter for reducing the effect of saturation; μ and v are preset thresholds; n is the number of pixels in the region Ω, s (x) represents the saturation at point x, and l (x) represents the luminance at point x;

the standardization and fusion processing methods are respectively as follows:

wherein,for normalizing luminance image, χ is a predetermined parameter for adjusting global intensity to avoid overexposure or underexposure of the image, η is a predetermined control L andthe mixing ratio parameter of (1);

step six, canceling image I_decPerforming gamma conversion to obtain L';

step seven, maintaining the initial defogged image J^LReplacing the luminance layer L of the image with L', and repeatingAnd forming a final defogged image.

2. The exposure compensation and edge enhancement method for a defogged image according to claim 1, wherein the atmospheric light A is estimated by: taking a dark channel I of a fog image I to be processed^darkTaking the pixel set B with the maximum brightness of 0.1%, and then taking the pixel set B corresponding to the pixel set B in the original image I_IFrom B_IAnd selecting three channels of RGB and the largest pixel point as the estimation of A.

3. The exposure compensation and edge-enhancement method of a defogged image according to claim 1, wherein an initial transmittanceThe estimation method comprises the following steps:

<mrow> <mover> <mi>t</mi> <mo>~</mo> </mover> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mn>1</mn> <mo>-</mo> <mi>&amp;omega;</mi> <munder> <mi>min</mi> <mrow> <mi>c</mi> <mo>&amp;Element;</mo> <mo>{</mo> <mi>r</mi> <mo>,</mo> <mi>g</mi> <mo>,</mo> <mi>b</mi> <mo>}</mo> </mrow> </munder> <mrow> <mo>(</mo> <munder> <mi>min</mi> <mrow> <mi>y</mi> <mo>&amp;Element;</mo> <mi>&amp;Omega;</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> </mrow> </munder> <mo>(</mo> <mfrac> <mrow> <msup> <mi>I</mi> <mi>c</mi> </msup> <mrow> <mo>(</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> <msup> <mi>A</mi> <mi>c</mi> </msup> </mfrac> <mo>)</mo> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>

where ω is a constant close to 1 but less than 1, c is one of the color channels { r, g, b } of the fog image I; Ω is a local window area centered on x, y denotes the pixels in the Ω window, I^c(y) represents the value of the c-th color layer of the fog image I at y, A^cAnd the value of the c color layer of A is shown.